eval on multi-gpus

[masn1310]

I use the following examples given by your readme:
python3 -m accelerate.commands.launch \ --num_processes=8 \ -m lmms_eval \ --model llava \ --model_args pretrained="liuhaotian/llava-v1.5-7b" \ --tasks mme \ --batch_size 1 \ --log_samples \ --log_samples_suffix llava_v1.5_mme \ --output_path ./logs/
and I found that 8 GPUs using the same cases during inference. How can I run different examples on multiple gpus to save time? Or, is there something wrong with my command?